We used a core set of 475 accessions from the Korean World Rice Collection (KRICE), consisting of 421 cultivated and 54 wild accessions ( Supplementary Table 1 ). The cultivated accessions, collected worldwide by the National Gene Bank of the Rural Development Administration, Republic of Korea, were classified into three varietal types: landrace, bred, and weedy (Maung et al., 2021a; Maung et al., 2021b; Phitaktansakul et al., 2022). These 421 cultivated rice accessions were further classified into six groups based on their ecotype: temperate japonica (279 accessions), tropical japonica (26 accessions), indica (102 accessions), aus (9 accessions), aromatic (2 accessions), and admixture (3 accessions) (Kim et al., 2007). We determined the ecotypes of the different rice accessions by utilizing whole-genome resequencing data (Phitaktansakul et al., 2022). The data was used to construct a neighbor-joining tree using PHYLIP software (Felsenstein, 2004). The resulting tree was then classified based on the previously reported ecotypes of known accessions. In addition, 54 wild rice accessions obtained from the International Rice Research Institute were included in the study ( Supplementary Table 1 ). Weedy rice is an invasive and problematic type of rice that has evolved from de-domestication events from cultivated rice (He et al., 2017; Li et al., 2017). On the other hand, wild rice is a distinct species native to specific regions and is known for its unique grain characteristics (Roma-Burgos et al., 2021). All accessions were cultivated in an experimental field at the Plant Resources Department, Kongju National University (Yesan Campus), following standard crop management and cultural operations.

A total of 421 cultivated rice accessions were subjected to phenotypic screening for AG tolerance. To break dormancy, seeds from each accession were incubated at 30°C for one day, 48°C for three days, and then at 30°C for one day. After surface sterilization with a 1% sodium hypochlorite solution for 10 min, the seeds were washed with deionized distilled water thrice. Subsequently, the seeds were subjected to two treatments: normal germination (control) and flooded germination. Twenty-one days is a commonly used duration for AG evaluation under flooded soil conditions (Ismail et al., 2009; Ghosal et al., 2019; Kuya et al., 2019). However, in vitro studies have shown that this period can vary from 4-15 days depending on the purpose of investigation (Nghi et al., 2019; Islam et al., 2022; Thapa et al., 2022; Shanmugam et al., 2023). We selected a 14-day period to capture the period up to the emergence stage, which is crucial for evaluating seedling vigor and emergence, ultimately reflecting coleoptile elongation. Ten seeds of each accession were wrapped in wet absorbent filter paper for normal germination and placed in a plastic box for 14 days. For the flooded germination treatment, ten seeds from each accession were placed in a 15 ml conical tube filled with sterile water ensuring they were completely submerged, and incubated in the dark at 30°C for 14 days. The water levels of the absorbent filter paper and tubes were maintained throughout the experiment. The experiments were performed in triplicate (total 30 seeds) for both treatments, and the length of the coleoptiles was measured after 14 days of germination using an ordinary ruler. The flooding tolerance index (FTI) was calculated as the ratio of flooded coleoptile length (FCL) to normal coleoptile length (NCL). Analysis of variance (ANOVA) was performed using SPSS software to evaluate the effects of genotype and genotype × environment interactions on the trait of interest. Sources of variation were determined to assess the coefficient of variation (CV) and estimate the relationship between NCL and FCL.

We utilized resequencing data from the Korean rice collection, which was generated using an Illumina HiSeq 2500 Sequencing System Platform. The cultivated Korean accessions were resequenced, with an average coverage depth of 13.5× ( Supplementary Table 1 ). Among the cultivated accessions, 327 had depths greater than 10× ( Supplementary Figure 1A ). For wild accessions, the average sequencing depth was 72.2×, with a maximum depth of 143.6×, and 46 accessions had depths greater than 40× ( Supplementary Figure 1B ). The resequencing data underwent a series of processing steps, including data preparation, filtering, mapping, sorting, and variant calling. The raw data was provided in FastQ format, and missing values were removed using VCFtools (Danecek et al., 2011). BWA v0.7.15 and Samtools v1.3.1 were used to index and align the Nipponbare reference genome (IRGSP 1.0) (Li and Durbin, 2009; Li et al., 2009). Duplicate reads aligned at multiple locations were removed using PICARDv1.88 (Toolkit, 2019). Final alignment and variant calling were performed using GATK tools v4.0.1.2 (McKenna et al., 2010). The resulting variants were filtered with VCFtools to eliminate false-positive SNPs/InDels. Default settings were used for most of the software and tools employed in the analysis. The VCF file containing information of genetic variants was used to evaluate genetic variation in the OsTPP7 gene. The sequencing data of the OsTPP7 gene were deposited in the NCBI GenBank database with the accession numbers MZ682675-MZ683149 ( Supplementary Table 1 ).

To assess the population structure among the different groups of rice accessions, we generated a PCA plot based on two principal components (PC1 and PC2) using TASSEL5 (Bradbury et al., 2007). The resulting variable components were visualized using the ggplot2 package in R (Wickham et al., 2016). Multidimensional scaling was employed to differentiate varietal groups (bred, weedy, landrace, and wild) based on various variables obtained from the VCF file for the OsTPP7 gene. We used VCFtools to convert the previously called variants into plink output using the PLINK analysis toolset (Purcell et al., 2007). This conversion generated a bed file, and two additional files (.bim and.fam format) were generated using a Python script. To investigate population structure, we utilized the FastStructure package tool (Raj et al., 2014). To explore population ancestry, we estimated the number of subpopulations ranging from 2 to 7. The admixed patterns of defined populations (i.e., population structure) were analyzed using average Q-values obtained from the Pophelper analytical tool (Francis, 2017) in RStudio.

To study the evolutionary relationship of the OsTPP7 gene among various plant species, we downloaded function-annotated protein sequences from 19 different species, available at (http://plants.ensembl.org/info/data/ftp/index.html) ( Supplementary Table 2 ). These species included 12 wild rice species: Oryza barthii, Oryza glaberrima, Oryza longistaminata, Oryza glumipatula, Oryza nivara, Oryza rufipogon, Oryza sativa japonica, Oryza sativa indica, Oryza sativa japonica, aus, Oryza meridionalis, Oryza punctata, and Oryza brachyantha, six cereals and one dicot. We identified orthologous genes by running OrthoFinder v2.5.4 (Emms and Kelly, 2015) and generated a neighbor-joining tree using MEGA7 (Kumar et al., 2016).

To analyze evolutionary patterns, we calculated nucleotide diversity (π), Tajima’s D, and population differentiation (FST) using VCFtools 0.1.13. First, we extracted variant files for the gene region of each group using VCFtools and compared them for both SNPs and InDels. Only sites with a minor allele frequency (MAF) greater than 0.05 and no missing data were included in the analysis. For nucleotide diversity and Tajima’s D tests (Tajima, 1989), we used a sliding window size of 1.0 kb and compared the results among the different groups of the 475 rice accessions. FST was also calculated using VCFtools v0.1.15, with a 500-bp slide window and 500-bp steps, to assess genetic differences among the groups.

Furthermore, we performed a haplotype analysis on the OsTPP7 gene region to group its functional variations (SNPs and InDels). VCFtools was used to create a FASTA file from the VCF file, and sequences were aligned using MEGA7 (Kumar et al., 2016). The aligned sequences were then transformed to the nexus format before conducting haplotype analysis with DnaSP v6.12.03 (Rozas et al., 2017). We constructed a TCS haplotype network (Crandall et al., 2000) using the Population Analysis with Articulate Tree (PopART) v1.7 software (Leigh and Bryant, 2015). Additionally, we conducted haplotype analysis in the 3K_RG panel for the OsTPP7 gene. The sequence data for the 3K_RG panel was downloaded from the online database https://snp-seek.irri.org/_download.zul;jsessionid=79B65E769E27730C020C4C198AE67235 accessed on May 22, 2020, and haplotyping was performed without filtering the variants.

All statistical analyses were performed using R Statistical Software (version 4.2.3; R Foundation for Statistical Computing, Vienna, Austria). The results are presented as the mean ± standard error (SE). One-way ANOVA and Sheffe’s test were used to detect significant differences. Further statistical comparisons were conducted to determine the associations between the identified functional haplotypes and AG tolerance using the phenotypic data from the cultivated accessions. The major haplotypes were used for comparison, and the mean phenotype was calculated for each group of accessions carrying a similar haplotype. The difference between the means was compared by the Student’s t-test using the ‘t.test’ function in the R.

The cultivated rice accessions from the Korean collection showed considerable variation in AG phenotyping, as summarized in Table 1 . Under normal conditions, the NCL ranged from 0.42 cm to 3.85 cm, with a mean of 1.94 cm. Under flooded conditions, the FCL ranged from 0.69 cm to 5.37 cm, with a mean of 2.80 cm. Japonica accessions displayed higher coleoptile lengths (CLs) than indica accessions under both normal and flooded conditions. The average CLs for indica, temperate japonica, and tropical japonica ecotypes under normal germination were 1.92 cm, 2.07 cm, and 1.99 cm, respectively. These values increased in response to AG, resulting in average CLs of 2.26 cm, 3.02 cm, and 2.89 cm, respectively. Consequently, the FTI values were also higher in japonica compared to indica ( Table 1 ).

The ANOVA revealed that the interaction effect between genotype and environment strongly influenced coleoptile lengths ( Table 1 ). Additionally, the genotypic effect was significant for all traits, while the environmental effect was not significant. The coefficient of variation (CV) for FTI was 46.23%, higher than that for NCL (30.02%) and FCL (32.42%). The standard deviation was also higher for FCL (0.90) compared to NCL and FTI ( Table 1 ).

Furthermore, we categorized the distinct variations in coleoptile traits into four groups based on length: short (<1.5 cm), intermediate (1.5-2.5 cm), long (2.5-3.5 cm), and very long (> 3.5 cm) ( Figure 1 ; Supplementary Figure 2 ). Interestingly, the number of accessions in the short and intermediate categories decreased, while the number in the long and very long categories increased under AG. Under normal conditions, 68 indica accessions were grouped in the intermediate category, but only 44 remained in that category under AG. Whereas the number of accessions in the long category increased from 10 to 36 under normal and AG conditions, respectively. Similar trends were observed for temperate japonica and tropical japonica accessions. While only three accessions had NCL >3.5 cm, 81 accessions showed very long FCL under AG ( Figure 1 ).

Regarding the different ecotypes, no significant differences were found in NCL. However, significant variations were observed in FCL and FTI between the ecotypes. Notably, both temperate japonica and tropical japonica ecotypes exhibited significantly higher FCL and FTI values compared to the indica and aus ecotypes ( Figure 2 ). These findings underscore the differential response of temperate japonica and tropical japonica ecotypes towards AG, as evidenced by their significantly higher FCL compared to the indica ecotypes.

The OsTPP7 gene (Os09g0369400) is located on chromosome 9 at 12,251,875.12,254,061 (+ strand) and consists of seven exons with a length of 2186 bp. Table 2 presents the genetic variants, including SNPs, insertions (Ins), and deletions (Del), in the OsTPP7 gene region of the 475 rice accessions. Among the cultivated group, temperate japonica had 8 SNPs, followed by six SNPs in indica. In contrast, no genetic variation was observed in the aromatic group, which matches the Nipponbare reference sequence. A total of 163 genetic variations (53 Ins, 33 SNPs, and 23 Del) were identified in wild rice.

The genetic composition of the Korean rice collection was further examined for OsTPP7 gene region using population structure and PCA ( Figure 3 ). It was observed that the indica and japonica ecotypes exhibited a somewhat similar structure at K = 3, K = 4, and K = 7. Interestingly, the indica ecotype was divided into two clusters, representing temperate and tropical japonica, indicating a certain degree of genetic similarity between indica and one or both of these ecotypes in relation to the OsTPP7 gene. In contrast, wild rice exhibited a mixed structure that became more apparent with increasing K values ( Figure 3A ). Similarly, in the PCA, the wild rice accessions exhibited a scattered distribution (part a in Figure 3B ), while some overlap was observed among the cultivated groups (part b in Figure 3B ). Furthermore, PCA analysis based on varietal type also revealed a similar dispersion pattern for bred, landrace, weedy, and wild accessions ( Supplementary Figure 3A ).

Our analysis of orthologous genes revealed distinct sub-clades formed by all TPP gene members, with TPP7 genes clustered together with TPP6 ( Supplementary Figure 4 ). Furthermore, in the PCA analysis, we observed clear differentiation of wild rice accessions from other ecotypes based on the first and second principal components (PC1 and PC2) ( Figure 3B ), which aligns with previous studies (Veasey et al., 2011; Huang et al., 2012). These findings suggest that wild rice may possess a larger genetic distance and a different genetic background, indicating a complex domestication history during the evolution of wild rice (Konishi et al., 2006; Choi et al., 2019).

To assess the level of differentiation among the populations, we calculated fixation index (FST) values based on the OsTPP7 gene using weighted methods (Weir and Cockerham, 1984). The resulting statistics were then analyzed for pairwise comparisons ( Table 3 ). The highest weighted average FST (0.533) was observed between the temperate japonica and wild groups, while the lowest was observed between the temperate japonica and aromatic groups ( Table 3 ). Interestingly, the wild group showed a relatively closer genetic relationship to the aromatic and admixture groups.

For the indica group, the highest pairwise FST value was observed with wild type (0.271), followed by temperate japonica (0.261), while lower genetic differentiation was observed for the aromatic and admixture groups. The FST values between the tropical japonica and admixture groups (0.318) and between the temperate japonica and admixture groups (0.274) indicated a higher degree of genetic differentiation among the cultivated population ( Table 3 ).

Regarding varietal types, FST values ranged from 0.003 (between bred and weedy) to 0.459 (between bred and wild). These results revealed that pairwise FST values between cultivated groups were lower than those involving the wild group, indicating a closer genetic relationship for OsTPP7 within the cultivated varietal types ( Supplementary Figure 3B ).

The nucleotide diversity (π) for the OsTPP7 gene was analyzed to evaluate the degree of polymorphism among different groups within the Korean rice collection. The results revealed that the wild group had the highest diversity, followed by admixture, indica, aus, temperate-japonica, and tropical-japonica groups ( Figure 4A ). On average, the nucleotide diversity was 0.00008 in tropical japonica and 0.00265 in wild rice ( Figure 4B ). Notably, the highest value of nucleotide diversity (0.00515) was observed at position 12,254,000 in the wild group, while the weedy group had the lowest value (0.00042) at the same position ( Supplementary Figure 5A ). When considering the mean nucleotide diversity values, the wild group had the highest value (0.00310), followed by landrace (0.00219), bred (0.00131), and weedy (0.00086) ( Supplementary Figure 5B ). The diversity of japonica in this gene region was lower than that of indica and other groups, supporting the hypothesis of selection for OsTPP7 during the domestication of japonica rice.

To gain further insights into the presence of selection and/or demographic changes within the population, we analyzed Tajima’s D values for the OsTPP7 gene and assessed the differences between the expected and observed segregation numbers due to selection. Here, indica type showed a positive value (0.97190), followed by the admixture (0.6003) and tropical japonica (0.21739) ecotypes, while negative values were observed in temperate japonica, aus, and wild types ( Figure 5 ). For varietal types, the average Tajima’s D values ranged from −0.94356 (wild) to 0.00020 (landrace), with only the landrace showing a positive value (0.00020) ( Supplementary Figure 6 ). The positive Tajima’s D values observed in the indica, admixture, and tropical japonica ecotypes suggest the presence of balancing selection or population contraction. In contrast, the negative Tajima’s D values observed in the temperate japonica, aus, and wild types suggest a deficiency of intermediate-frequency variants, which could be due to purifying selection or population expansion.

In order to investigate the genetic diversity and relationships among different haplotypes of the OsTPP7 gene, a haplotype network was constructed ( Figure 6A ). A total of 92 polymorphic sites, including 34 SNPs and 58 InDels, were detected within the OsTPP7 genic region. These variations were distributed across different regions, with 23 sites in the exon region, 43 sites in introns, four sites in the 5′UTR, and 22 sites in the 3′UTR ( Supplementary Table 3 ). The haplotype analysis revealed 18 distinct haplotypes, with three specific to cultivated accessions, 13 specific to wild accessions, and two present in both cultivated and wild accessions. The most common haplotype (Hap_1) was found in 395 rice accessions, including 362 cultivated and 33 wild accessions ( Supplementary Table 4 ). Notably, among 305 japonica accessions, 300 were japonica accession into Hap_1, while Hap_2 and Hap_3 were predominated associated with indica accessions ( Figure 6 ). A closely connected network was observed among the haplotypes of cultivated rice accessions, in which Hap_2, Hap_4, and Hap_5 were derived from the major haplotype, Hap_1, suggesting their close relationship ( Supplementary Figure 7 ). Conversely, the wild rice haplotypes formed a network with varying degrees of mutational steps, clearly demonstrating a considerable genetic distance between cultivated and wild rice ( Supplementary Figure 7 ).

Within the coding region of OsTPP7, we identified ten functional (non-synonymous) SNPs (referred to as fSNPs hereafter). One of these fSNPs was a C/A substitution at position 12,253,191 in exon 5, resulting in serine to arginine change found in Hap_2. This haplotype was present in 50 accessions, including 44 indica, one aus, two admixture, and three wild accessions. Another haplotype, Hap_3, was detected in six indica and four temperate japonica accessions. Hap_3 was characterized by a non-functional C/T SNP at position 12,252,673 in exon 3, a functional C/A SNP at position 12,253,191 in exon 5, and a 20-bp insertion in the 3’ UTR region. Additionally, Hap_5, found in one temperate japonica accession, had an fSNP G/A at position 12,252,716, causing a substitution from glycine to arginine in exon 3. The most prevalent haplotype, Hap_1, was present in temperate japonica (274 accessions), indica (52 accessions), aus (seven accessions), wild (33 accessions), admixture (one accession), tropical japonica (26 accessions), and aroma rice (two accessions) ( Figure 6 ).

Among the wild group, haplotype analysis identified 15 haplotypes, nine of which contained fSNPs. Hap_10 had one heterozygote functional allele, while the remaining eight fSNPs (including one C/A, three G/A, two C/G, and two G/C) were unique to the wild type. Hap_9, present in three wild accessions, had two fSNPs (C/G and G/C), while Hap_18, found in two accessions, had a G/A fSNP causing a valine to methionine substitution in exon 4 ( Figure 6 ).

Furthermore, we analyzed 3,000 rice genome (3K-RG) data to evaluate the OsTPP7 gene polymorphisms in a large number of accessions. A total of 112 haplotypes were identified, consisting of 105 polymorphisms, including 100 SNPs and 5 InDels. Among these, 72 SNPs were fSNPs located in the coding region, while the 5 InDels (two insertions and three deletions) were present in the intron and 3’ UTR ( Supplementary Tables 5 , 6 ). The primary haplotype Hap_1 was detected in 1,485 accessions, followed by Hap_2 in 1,107 accessions. Hap_2 was characterized by a C/A fSNP at position 12,253,191, causing a serine to arginine substitution in exon 5. Another major haplotype, Hap_5, was found in 223 accessions and featured the fSNP G/T at position 12,252,094 in exon 1. Hap_7 exhibited a 20 bp insertion in the 3’ UTR region and was characterized by the fSNP C/A at position 12,253,191 and the non-functional SNP C/T at position 12,252,673 ( Supplementary Figure 8 ).

Next, we evaluated the phenotypic performance of the major haplotypes, Hap_1, Hap_2, and Hap_3, across cultivated accessions (362 accessions in Hap_1, 47 accessions in Hap_2, and ten accessions in Hap_3) ( Figures 7A–C ). By conducting pairwise t-tests at a significance level of 0.05, we observed significant differences in FCL between Hap_1 and the other haplotypes, Hap_2, and Hap_3 ( Figure 7B ). For FTI, significant differences were observed between Hap_1 and Hap_2/Hap_3, although the significance levels were relatively lower compared to FCL ( Figure 7C ). Hap_2, an indica-specific haplotype with 44 indica accessions, displayed lower FCL and FTI values than Hap_1. However, there was no significant variation in NCL among the haplotypes ( Figure 7A ).

To further validate the findings from the conical tube experiment and provide additional evidence for the impact of OsTPP7 haplotypes on AG tolerance, we conducted a tray-based experiment for AG phenotyping. We randomly selected a total of 137 accessions from three major haplotypes, including Hap_1 (107 accessions), Hap_2 (23 accessions), and Hap_3 (six accessions). AG phenotyping was carried out based on the protocol of Septiningsih et al. (Septiningsih et al., 2013) ( Supplementary Methods ). The results revealed significant differences between Hap_1 and Hap_2/Hap_3 for FCL and shoot length ( Supplementary Table 7 ; Figures 7D, E ). These findings provide further support for the notion that allelic variations within OsTPP7 contribute to the phenotypic variation in coleoptile response to anaerobic germination rather than normal germination.